## Hückel-Theorie ### Von HF zurück zu LCAO ###################################################### LCAO für H$_2^+$ - ein Elektron, $\mathbf{H}$ fest - einmal diagonalisieren $$\mathbf{H}\,\mathbf{c} = E\,\mathbf{S}\,\mathbf{c}$$ Hartree-Fock - alle Elektronen, $\mathbf{F}(\mathbf{c})$ - selbstkonsistent (SCF) $$\mathbf{F}(\mathbf{c})\,\mathbf{c} = \mathbf{S}\,\mathbf{c}\,\boldsymbol{\varepsilon}$$ -|- Idee - $\mathbf{F}$ parametrisieren - kein SCF - nicht allgemein gültig Zwei Schritte - Parameter für $F_{\mu\nu}$ $\Rightarrow$ Extended Hückel[^1] - nur $\pi$, $\mathbf{S} = \mathbf{1}$ $\Rightarrow$ Hückel[^2] [1] Hoffmann, 1963. [2] Hückel, 1931. ### Fock-Matrix in Löwdin-Basis AOs nicht orthogonal $\Rightarrow$ $F_{\mu\nu}$ schwer vergleichbar Löwdin-Orthogonalisierung $$\tilde{\mathbf F} = {\mathbf{S}^{-1/2}}\,{\mathbf{F}}\,\mathbf{S}^{-1/2}$$ [[ \tilde{\mathbf F} ]] Löwdin-Basis [[ {\mathbf{S}^{-1/2}} ]] Orthogonalisierung [[ {\mathbf{F}} ]] AO-Basis $$\tilde{\mathbf{F}}\,\tilde{\mathbf{c}} = \tilde{\mathbf{c}}\,\boldsymbol{\varepsilon}$$ [[ \tilde{\mathbf{c}} ]] $\mathbf{S}^{1/2}\mathbf{c}$ [[ \boldsymbol{\varepsilon} ]] unverändert -|- Löwdin-Orbitale - orthonormal - jedes AO möglichst wenig verändert - weiterhin einem Atom zugeordnet Matrixelemente - $\tilde F_{\mu\mu}$: Energie eines Elektrons in AO $\mu$ - $\tilde F_{\mu\nu}$: Kopplung zwischen AO $\mu$ und $\nu$ - gewöhnliches Eigenwertproblem ### Ethylen: Fock-Matrix RHF/STO-3G, $\tilde{\mathbf{F}}$ sortiert nach $\sigma$- und $\pi$-AOs ```plot: from pyscf import gto, scf def lowdin_fock(th): c, s = np.cos(np.radians(th)), np.sin(np.radians(th)) mol = gto.M(atom=f'''C 0.667 0 0; C -0.667 0 0 H 1.23 0.92 0; H 1.23 -0.92 0 H -1.23 {0.92*c} {0.92*s}; H -1.23 {-0.92*c} {-0.92*s}''', basis='sto-3g', verbose=0) mf = scf.RHF(mol).run() e, U = np.linalg.eigh(mf.get_ovlp()) X = U @ np.diag(e**-0.5) @ U.T Ft = X @ mf.get_fock() @ X lab = [l.strip() for l in mol.ao_labels()] pi = [lab.index('0 C 2pz'), lab.index('1 C 2pz' if th == 0 else '1 C 2py')] sig = [i for i in range(len(lab)) if i not in pi] o = sig + pi return Ft[np.ix_(o, o)], len(sig) fig, axes = plt.subplots(1, 2) for ax, th, title in zip(axes, [0, 90], ['planar', r'verdrillt $90^\circ$']): Ft, ns = lowdin_fock(th) im = ax.imshow(Ft, cmap='RdBu_r', vmin=-0.8, vmax=0.8) for p in (ns - 0.5,): ax.axhline(p, color='k', lw=1) ax.axvline(p, color='k', lw=1) n = len(Ft) ax.set_xticks([ns/2 - 0.5, ns + 0.5], [r'$\sigma$', r'$\pi$']) ax.set_yticks([ns/2 - 0.5, ns + 0.5], [r'$\sigma$', r'$\pi$']) ax.tick_params(length=0) ax.set_title(title + rf', $\tilde F_{{\pi_1\pi_2}} = {Ft[-1, -2]:.2f}$') ``` ### $\pi$-Block von Ethylen $\pi$-Block allein: $2\times 2$ $$\tilde{\mathbf{F}}_\pi = \begin{pmatrix} {\alpha} & {\beta} \\ \beta & \alpha \end{pmatrix} \quad\Rightarrow\quad \varepsilon_\pm = \alpha \pm \beta$$ [[ {\alpha} ]] $\tilde F_{\mu\mu}$ eines $p_z$ [[ {\beta} ]] $\tilde F_{\mu\nu}$ zweier $p_z$, $\beta < 0$ - gleiche Mathematik wie H$_2^+$ - $\beta$ hängt am Überlapp der $p_z$ - $\pi$-Orbitale aus $2\times 2$ statt $14\times 14$ -|- ```plot: from pyscf import gto, scf def pi_block(th): c, s = np.cos(np.radians(th)), np.sin(np.radians(th)) mol = gto.M(atom=f'''C 0.667 0 0; C -0.667 0 0 H 1.23 0.92 0; H 1.23 -0.92 0 H -1.23 {0.92*c} {0.92*s}; H -1.23 {-0.92*c} {-0.92*s}''', basis='sto-3g', verbose=0) mf = scf.RHF(mol).run() e, U = np.linalg.eigh(mf.get_ovlp()) X = U @ np.diag(e**-0.5) @ U.T Ft = X @ mf.get_fock() @ X lab = [l.strip() for l in mol.ao_labels()] p1 = np.eye(len(lab))[lab.index('0 C 2pz')] p2 = np.zeros(len(lab)) p2[lab.index('1 C 2py')], p2[lab.index('1 C 2pz')] = -s, c P = np.array([p1, p2]) eps = np.linalg.eigvalsh(P @ Ft @ P.T) nocc = mol.nelectron // 2 return eps, mf.mo_energy[nocc - 1:nocc + 1] ths = np.linspace(0, 85, 18) res = [pi_block(t) for t in ths] blk = np.array([r[0] for r in res]) hf = np.array([r[1] for r in res]) fig, ax = plt.subplots() ax.plot(ths, hf[:, 0], 'o', ms=3, color='C0', label=r'HF $\pi$ (HOMO)') ax.plot(ths, hf[:, 1], 'o', ms=3, color='C1', label=r'HF $\pi^*$ (LUMO)') ax.plot(ths, blk[:, 0], '-', color='C0', label=r'$\alpha + \beta$ aus $2\times 2$') ax.plot(ths, blk[:, 1], '-', color='C1', label=r'$\alpha - \beta$ aus $2\times 2$') ax.set_xlabel(r'Torsionswinkel [$^\circ$]') ax.set_ylabel(r'$\varepsilon$ [Hartree]') ax.legend() plt.tight_layout() ``` ### Übertragbarkeit All-trans-Polyene C$_2$H$_4$ bis C$_{12}$H$_{14}$, RHF/STO-3G ```plot: from pyscf import gto, scf from pyscf.geomopt.geometric_solver import optimize def polyene(n): pos = [np.zeros(3)] for i in range(n - 1): d = 1.34 if i % 2 == 0 else 1.46 a = np.radians(30 if i % 2 == 0 else -30) pos.append(pos[-1] + d*np.array([np.cos(a), np.sin(a), 0])) atoms = [('C', p) for p in pos] for i, p in enumerate(pos): nb = [pos[j] for j in (i - 1, i + 1) if 0 <= j < n] if len(nb) == 2: v = 2*p - nb[0] - nb[1] atoms.append(('H', p + 1.09*v/np.linalg.norm(v))) else: u = (p - nb[0])/np.linalg.norm(p - nb[0]) w = np.array([-u[1], u[0], 0]) for sg in (1, -1): atoms.append(('H', p + 1.09*(0.5*u + sg*0.866*w))) mol = gto.M(atom=[(a, tuple(p)) for a, p in atoms], basis='sto-3g', verbose=0) mol = optimize(scf.RHF(mol)) mf = scf.RHF(mol).run() e, U = np.linalg.eigh(mf.get_ovlp()) X = U @ np.diag(e**-0.5) @ U.T Ft = X @ mf.get_fock() @ X pz = [i for i, l in enumerate(mol.ao_labels()) if 'C 2pz' in l] return Ft[np.ix_(pz, pz)] ns = np.arange(2, 13, 2) F = [polyene(n) for n in ns] fig, axes = plt.subplots(1, 2, sharex=True) axes[0].plot(ns, [f[0, 0] for f in F], 'o-', label=r'$\alpha$ endständig') axes[0].plot(ns[1:], [f[1, 1] for f in F[1:]], 's-', label=r'$\alpha$ innen') axes[0].set_ylim(-0.1, 0.1) axes[0].set_ylabel(r'$\tilde F_{\mu\mu}$ [Hartree]') axes[1].plot(ns, [f[0, 1] for f in F], 'o-', color='C2') axes[1].plot(ns[1:], [f[1, 2] for f in F[1:]], 's-', color='C3') axes[1].plot(ns[1:], [f[0, 2] for f in F[1:]], '^-', color='C7') axes[1].set_ylim(-0.4, 0.05) axes[1].set_ylabel(r'$\tilde F_{\mu\nu}$ [Hartree]') for ax in axes: ax.set_xlabel('Anzahl C-Atome') ax.set_xticks(ns) axes[0].legend() for y, t, c in [(0.01, 'übernächster Nachbar', 'C7'), (-0.176, 'lang (1.49 Å)', 'C3'), (-0.315, 'kurz (1.32 Å)', 'C2')]: axes[1].text(7, y + 0.03, t, color=c, ha='center') plt.tight_layout() ``` - unabhängig von Kettenlänge: gleiche Umgebung $\Rightarrow$ gleicher Wert ### Koopmans-Theorem Orbitalenergie in HF $$\varepsilon_i \approx -{\mathrm{IP}_i}$$ [[ \varepsilon_i ]] Orbitalenergie [[ {\mathrm{IP}_i} ]] Ionisierungsenergie - Elektron entfernen, übrige Orbitale eingefroren - Vernachlässigt: Relaxation, Korrelation (teilweise Fehlerkompensation) -|- Konsequenz für Parametrisierung - Diagonalelement $F_{\mu\mu}$: Energie eines Elektrons im AO $\mu$ - Atom-Ionisierungsenergien als Parameter ### Extended Hückel Parametrisierte Matrixelemente $$H_{\mu\mu} = -{\mathrm{VSIP}_\mu}$$ $$H_{\mu\nu} = {K}\,{S_{\mu\nu}}\,\frac{H_{\mu\mu} + H_{\nu\nu}}{2}$$ [[ {K} ]] empirisch, $K = 1.75$ [[ {S_{\mu\nu}} ]] Überlapp, berechnet - nur Valenz-AOs - $\mathbf{H}\,\mathbf{c} = E\,\mathbf{S}\,\mathbf{c}$, kein SCF - Kopplung $\propto$ Überlapp -|- Vorteile - ein Diagonalisierungsschritt - qualitativ richtige MO-Abfolge - alle Valenz-AOs Beschränkungen - keine Selbstkonsistenz: Ladungsumverteilung fehlt - keine verlässliche Gesamtenergie - keine Geometrien - Parameter atomweise, nicht molekülspezifisch ### Beispiel: Hexatrien Ausgangspunkt - $\tilde{\mathbf{F}}$ in Löwdin-Basis - planares Molekül - alternierend Doppel- und Einfachbindungen Näherungen - nur $\pi$: ein $p_z$ pro C-Atom, $\sigma$ entkoppelt - orthonormale Basis: $\mathbf{S} = \mathbf{1}$ - $\tilde F_{\mu\mu} = \alpha$ für alle C-Atome - $\tilde F_{\mu\nu} = \beta$ für gebundene Nachbarn - sonst $\tilde F_{\mu\nu} = 0$ -|- Hexatrien: $\pi$-Block von $\tilde{\mathbf{F}}$ (HF) ```plot: from pyscf import gto, scf from pyscf.geomopt.geometric_solver import optimize n = 6 pos = [np.zeros(3)] for i in range(n - 1): d = 1.34 if i % 2 == 0 else 1.46 a = np.radians(30 if i % 2 == 0 else -30) pos.append(pos[-1] + d*np.array([np.cos(a), np.sin(a), 0])) atoms = [('C', p) for p in pos] for i, p in enumerate(pos): nb = [pos[j] for j in (i - 1, i + 1) if 0 <= j < n] if len(nb) == 2: v = 2*p - nb[0] - nb[1] atoms.append(('H', p + 1.09*v/np.linalg.norm(v))) else: u = (p - nb[0])/np.linalg.norm(p - nb[0]) w = np.array([-u[1], u[0], 0]) for sg in (1, -1): atoms.append(('H', p + 1.09*(0.5*u + sg*0.866*w))) mol = gto.M(atom=[(a, tuple(p)) for a, p in atoms], basis='sto-3g', verbose=0) mol = optimize(scf.RHF(mol)) mf = scf.RHF(mol).run() e, U = np.linalg.eigh(mf.get_ovlp()) X = U @ np.diag(e**-0.5) @ U.T Ft = X @ mf.get_fock() @ X pz = [i for i, l in enumerate(mol.ao_labels()) if 'C 2pz' in l] Fp = Ft[np.ix_(pz, pz)] fig, ax = plt.subplots() ax.imshow(Fp, cmap='RdBu_r', vmin=-0.35, vmax=0.35) for i in range(n): for j in range(n): ax.text(j, i, f'{Fp[i, j]:.2f}'.replace('-0.00', '0.00'), ha='center', va='center') ax.set_xticks(range(n), [f'C{i+1}' for i in range(n)]) ax.set_yticks(range(n), [f'C{i+1}' for i in range(n)]) ax.tick_params(length=0) plt.tight_layout() ``` [*] Hartree, $\alpha$ als Nullpunkt bereits $\approx 0$ ### Struktur der Hückel-Matrix Hamilton-Matrix aus Bindungen $$\mathbf{H} = {\alpha}\,\mathbf{1} + {\beta}\,{\mathbf{A}}$$ [[ {\alpha} ]] Coulomb-Integral [[ {\beta} ]] Resonanzintegral [[ {\mathbf{A}} ]] Adjazenzmatrix $$E_k = \alpha + {x_k}\,\beta$$ [[ {x_k} ]] Eigenwert von $\mathbf{A}$ Welche Näherungen stecken in dieser Vereinfachung? ### Ethylen entspricht H$_2^+$ LCAO H$_2^+$, Basis $1s_A$, $1s_B$ - ein Elektron in $\sigma$ $$E_\pm = \frac{H_{AA} \pm H_{AB}}{1 \pm S}$$ $$\sigma \propto 1s_A + 1s_B \qquad \sigma^* \propto 1s_A - 1s_B$$ -|- Hückel Ethylen, Basis $p_{z,1}$, $p_{z,2}$ - $\alpha \leftrightarrow H_{AA}$, $\beta \leftrightarrow H_{AB}$, $S = 0$ - zwei Elektronen in $\pi$ - gleiche Mathematik, effektiver statt exakter Hamilton-Operator $$E_\pm = \alpha \pm \beta$$ $$\pi \propto p_{z,1} + p_{z,2} \qquad \pi^* \propto p_{z,1} - p_{z,2}$$ ### Hückel vs. HF $\pi$-Orbitalenergien: RHF/STO-3G gegen Hückel-$x_k$ ```plot: from pyscf import gto, scf from pyscf.geomopt.geometric_solver import optimize def chain(n): pos = [np.zeros(3)] for i in range(n - 1): d = 1.34 if i % 2 == 0 else 1.46 a = np.radians(30 if i % 2 == 0 else -30) pos.append(pos[-1] + d*np.array([np.cos(a), np.sin(a), 0])) atoms = [('C', p) for p in pos] for i, p in enumerate(pos): nb = [pos[j] for j in (i - 1, i + 1) if 0 <= j < n] if len(nb) == 2: v = 2*p - nb[0] - nb[1] atoms.append(('H', p + 1.09*v/np.linalg.norm(v))) else: u = (p - nb[0])/np.linalg.norm(p - nb[0]) w = np.array([-u[1], u[0], 0]) for sg in (1, -1): atoms.append(('H', p + 1.09*(0.5*u + sg*0.866*w))) A = np.diag(np.ones(n - 1), 1) + np.diag(np.ones(n - 1), -1) return atoms, A def ring(het): ang = np.pi/2 + 2*np.pi*np.arange(6)/6 r = 1.39*np.c_[np.cos(ang), np.sin(ang), np.zeros(6)] atoms = [('N' if het and i == 0 else 'C', p) for i, p in enumerate(r)] atoms += [('H', p*(1 + 1.08/1.39)) for i, p in enumerate(r) if not (het and i == 0)] A = np.roll(np.eye(6), 1, 0) + np.roll(np.eye(6), -1, 0) if het: A[0, 0] = 0.5 return atoms, A mols = {'Ethylen': chain(2), 'Butadien': chain(4), 'Hexatrien': chain(6), 'Benzol': ring(False), 'Pyridin': ring(True)} X_all, E_all = [], [] fig, ax = plt.subplots() for name, (atoms, A) in mols.items(): mol = gto.M(atom=[(a, tuple(p)) for a, p in atoms], basis='sto-3g', verbose=0) mol = optimize(scf.RHF(mol)) mf = scf.RHF(mol).run() pz = [i for i, l in enumerate(mol.ao_labels()) if '2pz' in l] w = np.einsum('mi,mn,ni->i', mf.mo_coeff[pz], mf.get_ovlp()[np.ix_(pz, pz)], mf.mo_coeff[pz]) eps = np.sort(mf.mo_energy[w > 0.5]) x = np.sort(np.linalg.eigvalsh(A))[::-1] ax.plot(x, eps, 'o', label=name) X_all += list(x) E_all += list(eps) b, a = np.polyfit(X_all, E_all, 1) xs = np.linspace(-2.2, 2.6, 2) ax.plot(xs, a + b*xs, 'k--', lw=1) ax.set_xlabel(r'Hückel $x_k$') ax.set_ylabel(r'HF $\varepsilon_k$ [Hartree]') ax.legend(loc='upper right') plt.tight_layout() ``` ### ? Zusammenfassung ################################################################## Extended Hückel - $F_{\mu\nu}$ parametrisiert, $\mathbf{S}$ berechnet - Diagonale aus Ionisierungsenergien - kein SCF -|- Hückel - nur $\pi$, $\mathbf{S} = \mathbf{1}$ - Bindungstopologie als Matrix - Ethylen: gleiche Mathematik wie H$_2^+$